CHARLES UNIVERSITY IN PRAGUE FACULTY OF SOCIAL SCIENCES Institute of Economic Studies Bachelor Thesis 2012 Ján Malega

Prepis

1 CHARLES UNIVERSITY IN PRAGUE FACULTY OF SOCIAL SCIENCES Institute of Economic Studies Bachelor Thesis 2012 Ján Malega

2 CHARLES UNIVERSITY IN PRAGUE FACULTY OF SOCIAL SCIENCES Institute of Economic Studies Ján Malega Countercyclical capital buffers in a new regulatory framework Bachelor Thesis Prague 2012

3 Author: Ján Malega Supervisor: PhDr. Jakub Seidler Academic year: 2011/2012

4 Bibliographic record Malega J. (2012): Countercyclical capital buffers in a new regulatory framework, Bachelor Thesis (Bc.) Charles University in Prague, Faculty of Social Sciences, Institute of Economic Studies, Prague, p., supervisor PhDr. Jakub Seidler.

5 Abstrakt Táto práca sa zaoberá vhodnosťou návrhu proticyklických kapitálových vankúšov ako novým nástrojom bankovej kapitálovej politiky. Koncept kapitálových vankúšov spočíva v tvorení dodatočných kapitálových rezerv v čase rastu ekonomiky, ktoré sa využívajú na stlmenie dopadov v období recesie. Práca rozoberá hlavné predpoklady a kľúčové premenné, ktoré pomáhajú určiť obdobie, tzv. nadmerného úverovania, teda čas, kedy by mali byť kapitálové rezervy uvoľnené. Konkrétne sa zameriava na metódu Hodrick-Prescott filtrovania, kde skúma, či metóda navrhnutá Basilejským výborom pre nastavenie kapitálových vankúšov je vhodná z pohľadu krajín strednej a východnej Európy, špeciálne v prípade Českej republiky. Za týmto účelom bol zavedený nový súbor ukazovateľov, ktorý určuje obdobie nadmerného rastu úverov v danej ekonomike. Na základe získaných výsledkov môžeme usúdiť, že vyhladzovací parameter navrhnutý Bazilejským výborom pre bankový dohľad použitý v Hodrick-Prescott filtri je v prípade konvergujúcich ekonomík nadhodnotený a môže viesť k nesprávnemu nastaveniu proticyklických kapitálových vankúšov v daných ekonomikách. Abstract This thesis discusses the relevance of the countercyclical capital buffer proposal as a new tool of bank capital policy. The concept of buffers is based on creating additional capital reserves at a time of economic growth, which are released in order to offset the impact in a recession. In this paper are discussed main assumptions and key variables that help to determine the excessive credit period, hence the time when should be the capital reserves released. Specifically it is focused on the method of Hodrick-Prescott filter, where examines, whether the method proposed by the Basel Committee for adjusting the capital buffers is appropriate from the perspective of Central and Eastern Europe, especially in the case of Czech Republic. For this purpose was introduced a new set of indicators that determines the period of excessive credit growth in the economy. Based on the results we conclude that the smoothing parameter suggested by the Basel Committee on Banking Supervision used in the Hodrick- Prescott filter in the case of converging economies is overestimated and may lead to improper adjustment of the countercyclical capital buffers in given economies.

6 Kľúčové slová: Basel III, nadmerné úverovanie, Hodrick-Prescott filter, ukazovateľ pomeru úverov ku HDP, krajiny strednej a východnej Európy Keywords: Basel III, excessive credit, Hodrick-Prescott filter, credit to GDP ratio, Central and Eastern Europe countries Volume: characters including spaces

7 Declaration of authorship 1. I hereby declare that I have written this bachelor thesis on my own under the guidance of my supervisor using only literature and other sources listed in reference list. 2. Furthermore, I declare that I have not used this thesis to acquire another academic degree. 3. I acknowledge and agree with lending and publishing of the thesis for study and research purposes. Prague, May 17, 2011 Ján Malega

8 Acknowledgement I would like to express my gratitude to PhDr. Jakub Seidler for his guidance and critique, as well as for his encouragement and patience as my supervisor.

9 Bachelor Thesis Proposal UNIVERSITAS CAROLINA PRAGENSIS založena 1348 Univerzita Karlova v Praze Fakulta sociálních věd Institut ekonomických studií Opletalova Praha 1 TEL: ,305 TEL/FAX: ies@mbox.fsv.cuni.cz Akademický rok 2010/2011 TEZE BAKALÁŘSKÉ PRÁCE Student: Obor: Konzultant: Ján Malega Ekonomie PhDr. Jakub Seidler Garant studijního programu Vám dle zákona č. 111/1998 Sb. o vysokých školách a Studijního a zkušebního řádu UK v Praze určuje následující bakalářskou práci Předpokládaný název BP: Countercyclical capital buffers in a new regulatory framework Předběžná náplň práce: Nedávna finančná kríza viedla k mnohým zmenám v politikách týkajúcich sa finančného sektora. Medzi nimi je vznik doporučení Bazilejskej komisie týkajúcej sa tvorenia dodatočných kapitálových rezerv, ktoré by banky mali zriadiť v časoch ekonomického rastu a následne ich využívať v období recesie. Tento koncept kapitálových rezerv by mal obmedziť proticyklickosť bankového sektora na hospodárskom cykle. Táto práca sa bude zaoberať opodstatnenosťou konceptu proticyklických kapitálových vankúšov ako nástrojom novej kapitálovej politiky. Budeme rozoberať hlavné predpoklady, empirickú evidenciu a predovšetkým to, kedy by malo byť pravidlo kapitálových vankúšov aktivované a následovne uvoľnené. Analýza bude spočívať v hľadaní vhodných premenných, na ktorých by mal byť proticyklický vankúš založený, najmä pre konvergujúce ekonomiky ako je Česká republika a zároveň overiť, či pomer úverov ku HDP je z pohľadu týchto krajín najvhodnejším indikátorom.

10 Předběžná náplň práce v anglickém jazyce: Recent Financial crisis led to many different changes in policies related to financial sector. Among them is the emergence of the Basel committee recommendations regarding to the creation of additional capital reserves that banks should set up in times of economic growth and subsequently use them in times of recession. This concept of capital buffers should limit countercyclicality of the banking sector on economic development. This thesis will discuss the validity of countercyclical capital buffers concept as the tool of new capital policy. We will discuss main assumptions, empirical evidence and mainly when should be capital buffers rule activated and when it should be afterwards released. The analysis will try to find appropriate variables on which should be countercyclical buffer based on, especially for converging economies such as the Czech Republic and whether the credit to GDP ratio is the most appropriate indicator for these countries. Seznam odborné literatury: DREHMAN, BORIO, GAMBACORTA, JIMÉNEZ, TRUCHARTE. Countercyclical capital buffers: exploring options, BIS working papers, ISSN , July 2010, available online at: BASEL COMMITTEE ON BANKING SUPERVISION. Countercyclical capital buffer proposal, Consultative document, ISBN , July 2010, available online at: EUROPEAN COMMISION. Consultation Document: Countercyclical Capital Buffer, February 2010, available online at: AYUSO, PÉREZ, SAURINA. Are capital buffers pro-cyclical? Evidence from Spanish panel data, Banco de España, revised October 2002, available online at: Trabajo/02/Fic/dt0224e.pdf STOLZ, WEDOW. Banks regulatory capital buffer and the business cycle: evidence for German savings and cooperative banks, Discussion paper, ISBN , July 2005, available online at: Datum zadání: Termín odevzdání: Podpisy konzultanta a studenta:. PhDr. Jakub Seidler. Ján Malega V Praze dne

11 Contents Introduction... 5 Literature review... 7 Excessive credit growth... 7 Problem of procyclicality... 8 Main features of an effective tool Approaches in identifying good and bad times Accumulating and releasing phase The taxonomy of possible plans Choosing the adjustment factor Methodology Hodrick-Prescott filter Problems of using Hodrick-Prescott filter Determination of parameter Hodrick-Prescott filter and CEE countries Calculation of countercyclical capital buffers I. Calculation of the aggregate private credit-to-gdp ratio II. Calculation of the credit-to-gdp gap III. Transformation of the credit-to-gdp ratio into a countercyclical buffer Disadvantages of Credit-to-GDP gap indicator Determination of the excessive level of credit Average credit growth indicator Average trend credit indicator Comparison of the results Bank capital to assets ratio Appropriate lambda for CEE countries Summary References: Appendix Appendix A: Appendix B: Appendix C: Appendix D:... 53

12 List of Graphs Graph 1: Graph 2: Graph 3: Graph 4: Graph 5: Graph 6: Graph 7: Graph 8: Graph 9: Procyclical assessment of credit risk Types of countercyclical capital buffers schemes Relationship between the countercyclical capital buffer and credit-to-gdp gap Comparison of GDP per capita Comparison of the Domestic credit to private sector United Kingdom, HP filter prediction Comparison Credit-to-GDP and its HP trend Comparison of the Gap and the Buffer Comparison of the credit-to-gdp growth Graph 10: Credit-to-GDP gap, Czech Republic until 2000 Graph 11: Credit-to-GDP gap, Czech Republic until 2012 Graph 12: Graph 13: Graph 14: The average credit growth indicator in United Kingdom The average trend credit indicator in United Kingdom Comparison of Bank capital to assets Graph 15: Credit-to-GDP gap, Czech Republic at λ= Graph 16: Adjusted level of λ for Czech Republic 2

13 List of Tables Table 1: Table 2: Table 3: Table 4: Table 5: Criteria to identify bad times Calculation of the countercyclical capital buffer for United Kingdom The Average credit growth indicator s results comparison for other countries The Average trend credit indicator s results comparison for other countries Calculation of excessiveness by the Average credit growth indicator and the Average trend credit indicator for United Kingdom List of Appendices Appendix A: Comparison of different level of λ for CEE countries until 2008 (graph) Appendix B: Appendix C: Comparison with countries which have experienced the consolidation (table) Relationship between the length of time series and value of parameter (graph) Appendix D: Computation of the countercyclical capital buffer with λ=20000 for Czech Republic (table) 3

14 List of Abbreviations BCBS BOE CEE CEMFI CEPR CNB CRSG CZK FSA HP IFS IMF MNB NBP NBS RWA WB Basel Committee on Banking Supervision Bank of England Central and Eastern Europe (countries) Centre for Monetary and Financial Studies (Spain) Centre for Economic Policy Research (Spain) Czech National Bank Credit Risk Standing Group (United Kingdom) Czech crown Financial Services Authority (United Kingdom) Hodrick-Prescott (filter) International Financial Statistics (IMF) International Monetary Fund Magyar Nemzeti Bank (Hungarian National Bank) National Bank of Poland National Bank of Slovakia Risk Weighted Assets World Bank 4

15 Introduction Recently we have experienced one of the greatest financial strain period also referred to as Global Financial Crisis. Previous exposure to the crisis has proved that regulation based on improving resilience of credit institutions is not sufficient. This incentive has started many attempts to create policy which will make current financial system more resistant. Objectives for the future are to analyze regarding factors and prevent the whole financial sector by limiting the systematic risk to ensure to not to repeat such an incident again. The Basel Committee introduced new capital adequacy rules which will be issued in Basel III regulatory standards and will deal with problem of procyclicality of the financial system (Borio et al., 2010: 1). These standards will be more strict, including improvement of Tier I and Tier II capital, higher capital requirements and introduction of the new policy proposal called countercyclical capital buffer aimed as protection of the banking system from times of excessive credit growth. The purpose of Basel III regulatory is primarily to reinforce the resilience of the financial system and promoting stability and sustainable economic growth. This should be obtained by better risk coverage, higher capital requirements and tighter capital definitions and also by building up the capital buffers. Basel III will be phased in period between Capital buffers should be introduced between , but the phase-in will likely be in this case accelerated (Harmsen, 2010: 98). The proposal of countercyclical capital buffers should encourage banks to build up buffers in financial good times in order to release them in case of bad times to ensure that credit institutions will be able to deal with stressful periods when are hit by high losses. If any of the institutions will fail to follow the capital buffer rules, they will be prohibited to distribute their profits as for example dividend payments (Repullo and Saurina, 2011: 5). The idea of buffers is to gradually increase capital requirements in case when credits expand faster than GDP growth. This instrument is supposed to prevent banks and make them more resilient in times of economic downturn. Relax the creating of credit bubbles and avoid the situation when banks have to raise new capital in the worst times. Main objectives of the buffers are to smooth the business and 5

16 financial credit cycle in financial sector and limit the risk of the losses by raising its resilience against shocks (Borio et al., 2011: 1 3). There are still many open questions about the features of the buffers. As the time of releasing, what means also correct identification of good and bad times; size of the buffer large enough to cover losses without making any other strains; international applicability and possibility to create automatic stabilizer in order to avoid manipulation. Also the cost of implementation (which should be low), transparency and simplicity of this framework should be considered (Borio et al., 2010: 1). In this thesis we examine these questions and we analyze suitable conditional variables which will help us to estimate the time of release of the buffer in the case of Central and Eastern European countries, especially Czech Republic. We discuss advantages and disadvantages of the Hodrick-Prescott filter method, consult other approaches and try to find out what is the most precise method in estimation of the credit excessiveness. Finally determine under what conditions is the credit-to-gdp ratio appropriate indicator for excessive credit growth in this type of countries by comparing our results with other approaches. The thesis is structured as follows. The second section, Literature review discusses the issues which can arise from excessive credit growth and the problem of procyclicality of economic sector and its impacts on given economy. The third part of this section is aimed on description of main features and approaches of an effective tool against the excessive credit expansion, discussing also suitable leading variables as indicators of potential credit booms. The third section Methodology introduces Hodrick- Prescott filter method also as its advantages and disadvantages. This chapter also presents different approaches in determining the smoothing parameter lambda and introduces our sample of the Central and Eastern European countries and the problems associated with them. Next section the Calculation of countercyclical capital buffers illustrates detailed step-by-step computation of the Credit-to-GDP indicator in case of United Kingdom as a good representative of this method and also discuss its drawbacks in case of CEE countries. The fifth chapter is devoted to introduction of new set of indicators and their features and also consults their results comparing them to Financial Stability Reports from the national banks of given countries. Following section The appropriate lambda for CEE countries analyzes the appropriateness of smoothing λ 6

17 parameter in case of Czech Republic by comparing various approaches and proposes a final suggestion. The summary attempts to describe obtained results and proposes other possible approaches in this topic. Literature review Excessive credit growth Rapid credit growth in some developing countries caused many concerns about its consequences to macroeconomic stability. Losses in the banking sector can be extraordinarily high when a depression is enhanced by preceding period of excessive credit growth (BCBS, 2010c: 57). Unreasonable high credit loans have significant impact on aggregate demand by indirect support of consumption in private sector. Problems which can arise from excessive credit growth are substantial, causing overheating of the economy; they influence indirectly inflation, interest rates and also real exchange rates. Changes in real exchange rates, especially its depreciation can also raise problems with credit loans on foreign exchange market and their financing from foreign sources. It will cause outflows of short-term foreign funds and flowingly inflating the credit bubble what can cause serious financial problems (Geršl and Seidler, 2011: 114). The reason of this risky lending behavior is called moral hazard. As we now with rising profitability is rising also profiteering and hence risky behavior of financial institutions thus their expectations about the future borrowers solvency are unreasonably optimistic. Especially in the case of too big to fail institutions, where their self-confidence is secured by reliance that government cannot afford to let them fall. The probability of having crisis after lending booms is about 75% higher comparing to normal times. Expectation to have banking crisis before a lending boom is slightly lower than during the normal times (Gourinchas et al., 2001: 14). Therefore many questions raised about the possible relation between credit growth and financial instability in that time. One reason for excessive credit growth in transition economies can simply imply their convergence in numbers to the advanced nations. However, it is 7

18 hard to decide what the exact definition of excessive credit growth in given countries is (Hilbers et al., 2005: 10). In the central banks and supervisory authorities proclaimed the situation as threatening and introduced series of special instruments for restricting excessive credit growth. There was a range of tools that includes from very soft until hard restrictions. For example an increase risk weights on selected loans or even restriction on credit portfolio growth. It is hard to discuss the efficiency of these instruments, mainly because they were released only short period before the global financial crisis raised. So the relevance of their usefulness is uncertain. Many studies which are interested about this problem claimed that before-mentioned instruments for restriction of credit growth are considerably ineffective (Geršl and Seidler, 2011: 113). These interactions emphasize the importance of building up appropriate capital defenses in times when risk in financial sector is growing remarkably (BCBS, 2010c: 57). Problem of procyclicality Financial procyclicality can be described as reciprocally strengthening interactions between the financial and real state of economy that incline to amplify the business cycle. This causes very often a financial volatility (Borio et al., 2011: 2). Procyclicality, in this context, is a term which refers to the tendency for regulatory capital requirements to rise with downswings in the economy and to fall with upswings. It is associated mainly with credit risk, as it is generally assumed that changes in operational risk are not correlated with the economic cycle. (FSA, 2004: 2, ) In 2004 was established second set of recommendations and regulations for banks known as Basel II. The idea was to achieve more risk-sensitive capital requirements in order to lower the incentives of excessive risk-taking and also help to protect system during the times of economic downswings. The goal was to reduce excessive risk-taking behavior and prevent lending in times when economy falls into a recession (Repullo et al., 2010: 105). 8

19 Even in the old Basel I regime of essentially flat capital requirements, bank capital regulation had the potential to be procyclical because bank profits may turn negative during recessions, impairing bank's lending capacity. Under the internal ratings-based (IRB) approach of Basel II, capital requirements are an increasing function of the probability of default (PD) and the loss given default (LGD) estimated for each borrower, and these parameters are likely to rise in downturns. So the concern about Basel II is that the worsening of borrowers' creditworthiness in recessions will increase the requirement of capital for banks and lead to a severe contraction in the supply of credit. (Repullo et al., 2010: 105, ) Graph 1: Procyclical assessment of credit risk (Source: Borio et al., 2011: 3) In the graph 1 we can see the procyclicality of the credit risk measure applied on hypothetical credit portfolio based on the risk-weights. Method used is called Internal rating based approach (IRB) we mentioned before. In the picture we can see the credit risk is lowest right before the crisis in 2007 and right after this point it shoot up rapidly. It is clear demonstration of mutual interactions between financial and real cycles (Borio et al., 2011: 3). This problem of procyclicality can considerably worsen the negative influence on bank s supply of credit and also whole economy during the time of economic downturns (Repullo et al., 2010: 106). A vulnerable financial system is no more able to 9

20 absorb losses without cutting down the risk and credit lending, what leads to fire sales and credit crunches (Borio et al., 2011: 2). Main difference between Basel I and Basel II is that Basel II is trying to strengthen the stability of banks, making them more credit-sensitive, therefore there is a much more lower probability of bank failures (Pausch and Welzel, 2011: 17). So we identify three main goals broadly speaking about countercyclical approaches. The most difficult is to smooth the business cycle by the capital requirements setting tool, in other words demand management tool. Second goal is to smooth financial or credit cycle and last one is to generally raise the resilience of the banks against the shocks without any changes in these two cycles (Borio et al., 2011: 3). Main features of an effective tool We spoke about the problems of procyclicality of the financial system and generally proposed some main objectives of countercyclical policies. The aim of countercyclical capital buffer is to ensure that banking sector is covered and has a sufficient capital buffer supposing to preserve system against the future potential losses (BCBS, 2010c: 57). More precise definition of what criteria have to be fulfilled in order to make our countercyclical capital buffer instrument effective is needed. Effectivity in this case means that this tool will be able to prevent banks against systematic risks by amplifying their defenses. So the releasing buffer will be in financially bad periods facilitating the credit supply strains. From this view we should focus on these four main criteria (Borio et al., 2011: 6): I. Timing: In order to make our tool the most effective, the exact amount has to be launched and released at proper time with the right speed. This implicate that we need to precisely define good and bad times (Borio et al., 2010: 5). II. Size of the buffer: We should ensure that size of the buffer accumulated will be sufficient to cover all losses in stressful times without making whole system to suffer from bigger strains. 10

21 III. IV. Robust to regulatory arbitrage: There should be no possibility to manipulate this tool by individual institution, also applicability of this instrument is expected to be abroad without any problems. Rule-based as possible, transparent, cost effective tool: In order to prevent it from manipulations and other things which can disrupt functionality. (Borio et al., 2011: 6) Let s describe these criteria more closely: Approaches in identifying good and bad times Good and bad times usually match the cycle of the economy thus its expansion and contraction stages. However the financial cycle in real economy is hard to measure. There are quite lot of variables which can be useful: Measures of bank performance: e.g. earnings, losses, various asset prices Financial activity: credit condition surveys Cost of availability of credit: credit spreads, funding spreads Real GDP growth Statistical data indicating ability of entities to meet their debt obligations on time (Borio et al., 2010: 5; BCBS, 2010b: 4) However looking for precise measures is very hard, for example measures for losses are hardly to get, mainly because accounting rules in differentiate across the countries and usually deform them. Also we often face the problem of lagging, what makes our measures often useless. Other possibility to measure bad times is to use historical data of banking crisis as proxy variable for this measure. Advantage of this approach is that there is very good historical database for all the countries, what can lead us to really good results, if the right method is used (Borio et al., 2011: 7). There is also other suggestion for another approach: Transition from good to bad times is in our case more critical, because most important is to release the buffer quickly enough and also in the sufficient amount. 11

22 This change could be identified by two factors: Normalized measure of aggregate gross losses Indicator which measures whether the banking sector is a source of credit contraction or not Table 1: Criteria to identify bad times Banking sector source of credit contraction Yes No High Bad times Bad times 1 Bank losses Low Bad times Good times (Source: Borio et al., 2010: 6) In order to determine these two measures we can use proxy variables. As appropriate proxy variable for aggregate gross losses we can use for example bank charge-offs. On the other hand proxy that indicates credit contractions will be harder to get, but we can still use some surveys (e.g. Senior Loan Officer Survey) (Borio et al., 2010: 7). Accumulating and releasing phase Suitable variables have to be find to provide sufficient guide how to accumulate buffers at proper speed and flowingly release them at a proper time. Promptly release of the buffer in bad times can lower the risk of the supply of credit, which is limited by the capital requirements. Also the proper authorities should inform about the time, how long they expect the release to last in order to help banks in future to lower their uncertainty about capital requirements. As BCBS (2010b) mentioned we can face several problems here: One of them could be that it is hard to acquire actual temperature of the good and bad times. The good leading variable which will estimate it and therefore release the buffer at proper timing and amount should be during the build-up phase copying its 1 Although are the bank losses high, credit supply is still not restricted, because banks are still trying to protect customer relations. 12

23 long-term average with the minimum variation. Ideal leading variable can be for example proxy for the build-up of the risk in good times (Borio et al., 2010: 8). Borio and Drehmann (2010) consider credit booms as the best single-variable leading indicator of banking distress, or even better is combination of credit and asset price deviations from long-term trends. Other problem comes with the question whether we can find single-variable which will be both, best leading and contemporaneous indicator of financial strains. If this variable exists, it will be exactly what we are expecting as the best indication point for the accumulation a release stage. There are a number of possible variables. We can divide them into three main groups and so: Macroeconomic variables, measures of banking activity and proxy variables for cost of funding (Borio et al., 2010: 9). Next section briefly describes adepts for the conditional variables. Macroeconomic variables: Between the most common macroeconomic variables belong measures of aggregate output, credit or asset prices. Advantage of using these variables in comparing with others is that there is no possibility for strategic manipulating or influence from individual institutions. Other advantage is the very good availability of these indicators all over the countries. On the other hand they can still be manipulated with the mutually controlled behavior of the banking sector. First most known basic variable for measuring the strength of given economy is Real GDP growth (Borio et al., 2010: 9). Real GDP growth: Measure which indicates the value of all goods and services produced by all residents in a given year, net of inflation, expressed in base-year prices (World Bank, ). However financial strains are not conditional to recessions, it means that they do not have to occur after every recession (Borio et al., 2010: 9). Aggregate real credit growth: This variable indicates real growth in supply of credit over some period from all of the sources, not only bank sector. This measure is interesting because expansion periods are usually accompanied by period of credit 13

24 booms and stressful times are reversely accompanied by decreasing of credit supply (Borio et al., 2010: 10). Asset price growth: Advantage of this estimator is that there is a tendency for these indicators to rise in time close before the systematic banking crises. Otherwise, in financially bad period they tend to decline rapidly. As an asset price growth we consider changes in specific property prices depending on economic cycle. As the indicator is usually used the difference of aggregate property prices from their longterm trend, also referred as asset price growth gap ratio (Borio et al., 2010: 10). This indicator is useful in predicting build-up phase. Although some problems can arise especially because deviations have tendency to narrow long time before financial strains emerge and this can cause releasing buffer earlier as needed. Nevertheless, the past of these indicators can be still useful in assessing the need to release the buffer in stressful times (BCBS, 2010b: 9). Credit-to-GDP ratio: This ratio is normalized credit ratio where we are also taking in account changes in demand and supply of credit with fluctuations in economy what is indicated by the GDP growth. Importance of this ratio lies in the fact that historically based during the crises was growth of the credit-to GDP faster than average growth. Same as we mentioned for asset price growth index, we use differences or deviations of credit-to GDP ratio from its long-term trend, also referred as credit-to-gdp gap (Borio et al., 2010: 10). This ratio acts very well in foreseeing the bad times going up fast before materialization the strains. It inclines to move up quickly above the trend before the serious episodes. Although neither credit-to-gdp ratio, nor aggregate real credit growth ratio are good indicators of the release phase, in words of timing or volume of the release. In many cases they just respond too slowly, creating lags what makes big problems in case of financial emergency. Last crisis is good example of it (Borio et al., 2010: 12). Banking sector activity: Indicators which measure activity in banking sector have tendency to be dependent on business and financial cycle. Variables like growth rate of lending, bank income or losses are measured more narrowly comparing with financial sector as whole and therefore they can tell us more specific data. Generally in periods of high bank 14

25 profitability we can usually see incentives from banks rising their credit activities even though they are facing a higher risk (Borio et al., 2010: 10). Bank credit growth: Measure growth of credit issued by banks, normalized by GDP (Borio et al., 2010: 10). Banking sector profits: Main indicator of given sector efficiency. Generally profits are high in financially good times and reversely they decline very fast in bad times. Although there is also possibility of manipulation from strategic management, so the numbers do not have to exactly show real image of current situation (Borio et al., 2010: 11). We can use this measure as predictor of build-up phase, however sometimes our results are unequal. This indicator is working very well in case of the United States or United Kingdom. Otherwise for Spain it performs poorly, probably because of different accounting practices (BCBS, 2010b: 9). Aggregate losses: This measure focuses on the cost part of banking sector. For example non-performed loans, provisions, write-offs or charge-offs and losses mostly associated with issuing credit loans. They use to fluctuate depending on current financial cycle. The reason for countercyclical capital buffers is exactly to eliminate these losses in stressful times by releasing the buffers accumulated in times when the losses are low (Borio et al., 2010: 11). However the performance of proxy variables for bank losses is not very precise in good times. Generally they fail to differentiate in financial good period because of absence of losses, what makes call for very high buffers very soon in the expansion (BCBS, 2010b: 9). Cost of funding This group of variables focuses on costs of banks related to raising funds. Thought is that banks should raise their funds in financially stabilized period when they are much more cheaper comparing to the stressful times when their price is rising significantly. Then they can use them later as reserves in these bad times in order to save money (Borio et al., 2010: 11). In last crises we faced, these measures acted very well. Signal is that they are below their long-term average and as strains are coming they tend to grow very quickly. Although, in measuring the multiple cycles, for example from data of the United States there is lack of precision (BCBS, 2010b: 9). 15

26 Banking sector credit spreads: Credit spread is difference between yields of various securities because of different quality. Spreads are indicators of vulnerabilities in financial or banking sector. In analysis are used averages of credit spreads from the biggest banks in each country (Borio et al., 2010: 11). Cost of liquidity: Measures of the average cost which has to be paid by banking sector when it wants to raise short-term liquidity. These indicators help us to see bank s health condition. While in good times there are no problems with distribution of liquidity by interbank markets in bad times we can feel how tension emerges. These measures could be also good indicators for identifying passing from good to bad times. On the other hand there is still space for strategic manipulating of these ratios because many of them like Libor are not transition-based measures, they are only based on agreement among these banks (Borio et al., 2010: 11). Corporate bond spreads: Is the difference between yields of corporate bonds comparing to government bonds. Their aggregate averages are used in analysis. It is the indicator of credit quality of the economy as a whole and as well as point-in-time measure of the credit risk. Generally the boom periods are accompanied by lower spreads then their average levels, but stressful periods are on the other hand characterized by fastwidening spreads. Also spreads can be seen as measure of the cost of borrowing in the economy, therefore it can be used as a tool that focuses on smoothing the costs of funding (Borio et al., 2010: 12). Conclusion Above mentioned possible indicator variables are further discussed in BCBS (2010b). The Credit-to-GDP gap ratio was the best performing indicator in this analysis. As we mentioned above, the analysis uses a specific form of this ratio also called Credit-to-GDP gap. It has many advantages over the Aggregate real credit growth measure. Expression as a GDP ratio, normalize our variable by the size of our economy we are working in. This also helps to annul the influence by the normal cyclical patterns of credit demand. The taxonomy of possible plans Any plan or scheme requires two components: Conditioning variable which tell us the time for build-up and release phase of the capital buffers. The first one we have 16

27 discussed above. The second is the adjustment factor that signalizes how to transfer changes in the conditional variable into capital variables, what are we going to analyze now (Borio et al., 2010: 2). There are two basic types of buffer creation: Minimum capital requirements: This approach includes countercyclical moving of the capital requirements, that is rising in good times or differently said in expansion stage and reversely falling in the release stage. When capital requirements are falling in the release stage the buffer is indirectly increasing, thus uprising by freeing capital. In order to have efficient buffer, speed of falling of capital minimum has to be higher than speed at which losses are incurred. Problem comes, when falling of the capital requirements in bad times drops under the current minimum, what violates fundamental constrains. Setting a target above the minimum: Setting target above minimum will create gap which moves counter-cyclically. Increasing in expansion phase and declining in contraction phase. There are more ways of setting targets. One extreme would be to set fixed buffer to be percent based on the minimum. We can see it in the graph 2., type 1. (Borio et al., 2010: 3). Graph 2: Types of countercyclical capital buffers schemes (Source: Borio et al., 2010: 3) We must take attention and try to avoid the inducing procyclicality in case of fixed instruments. 17

28 Another approach can be to increase target until any specific maximum in financially good times. Build-up phase should be connected to some conditioning variable as we discussed above (e.g. credit-to-gdp ratio). Question is whether to release the buffer instantly or gradually in stressful period. In the graph 2., type 2., we can see instant releasing phase, on the other hand graph 2., type 3. shows us a gradual releasing phase (Borio et al., 2010: 3). Choosing the adjustment factor After choosing the conditioning variable we need to link it with its adjustment factor what determine the way of building-up and releasing of the capital buffers required. There are more ways how to choose them. The adjustment factors are multiplicative and never lower than one in order to prevent the capital from dropping below the minimum capital level required (Borio et al., 2010: 19). Theoretically the simplest functional form when we take in account constrains will look like this: Graph 3: Relationship between the countercyclical capital buffer and credit-to- GDP gap (Source: Repullo and Saurina, 2011: 7) Where the on vertical axis represents the buffer as percentage of RWA and on horizontal axis is a credit-to-gdp gap. 18

29 The main objective of the countercyclical capital buffers is to protect banks from times of excessive credit growth. There are two important thresholds, the lower threshold usually marked as L, that starts the building phase and upper one, marked H, that limits the maximum of the buffer. The lower L and upper H threshold levels are key in determining the correct timing and the speed of releasing of the buffers (BCBS, 2010b: 13 16). When the conditioning variable is under the level L, the adjustment factors is equal to one. This is connected with bad times. Reversely in good times, when the conditioning variable is higher than L. Function itself is linear to the conditioning variable (Borio et al., 2010: 20). Criteria for determining lower and upper thresholds: Minimum threshold (L) level should be on one hand low enough, to ensure that buffer will start to accumulate sufficient time before a potential crisis, on the other hand it also should be high enough, to not to provide additional not required capital in normal times. While maximum threshold (H) upper level should be low enough to ensure that the buffer will be available on its maximum prior to major crises (BCBS, 2010a: 26). Recommended values of thresholds by BCBS are not binding for all monetary authorities. They provide just simple guide to determine them correctly, but their setting depends entirely on monetary authorities in given country (BCBS, 2010b: 16). As the whole topic is very broad, in this thesis we will focus only on the method of determining the period of excessiveness. Methodology Hodrick-Prescott filter The purpose of the research of Hodrick-Prescott filter (later HP filter) was to document main properties of aggregate economic fluctuations, also known as business cycle. It is based on hypothesis that growth component of aggregate economic time series deviate over the time. Movements of cyclical components are significantly different comparing to movements of the corresponding variables in macroeconomic 19

30 time series. HP filter is a tool designed to detrend given time series data, thus separate their cyclical component. Equation used in this filter is: [1] Where given time series is sum of a growth component and a cyclical component for all First part of the equation is the sum of the squared deviations. Second part is the sum of the second differenced squares of the trend components, multiplied by the parameter. The parameter is a positive constant that works as a smooth adjustment of the trend. The larger the is, the smoother is the resolution of the series (Hodrick and Prescott, 1997: 1 3). Problems of using Hodrick-Prescott filter Time series data used in computation of a trend by HP filter are significantly dependent on the length of given time series and also on the parameter lambda which has to be selected as well. This can expose us to end-point bias problem, which produces a highly uncertain estimation at the end of the data period. There is way to deal with it by extending the time series by its prediction in to the future, what on the other hand can cause even more uncertainty and therefore even worse our estimation and explanation of macro prudential policy (Geršl and Seidler, 2011: 115). Authorities in every country should take into account also other variables and compare them whether are consistent with the results of the credit-to-gdp ratio. Good examples of useful indicators are financial or total assets in private sector, funding or CDS spreads, real GDP growth or credit condition surveys. We can consider also any proxy variable for solvency and others which could be also appropriate substitutes for measuring the excessive credit growth (BCBS, 2010a: 8). Other significant criticism of HP filter is that this method does not include any economic fundamentals which affect the equilibrium stock of loans. There is another approach to measure and estimate the equilibrium of private credit level related to key economic variables, for example the level of development of given economy. As proxy variable we can use the real GDP per capita as some standard for living. This implicates that countries with the same level of economy or development should be on 20

31 the same level of their equilibrium for the private credit. Proportionally poorer countries should have lower levels of equilibrium than rich countries generally said with higher level of development. As some analyzes show, this is not always fulfilled (Geršl and Seidler, 2011: 116). Determination of parameter There are more approaches in technical literature about the setting lambda parameter. Papers from Ravn and Uhlig (1997) suggest to set lambda according to the expected duration of the cycle and frequency of its observations. Hodrick and Prescott (1997: 7) recommend to anchor lambda for quarterly data at level, what become standard indicator for business cycles. Analysis has shown that length of quarterly time series in business cycles should be around 7,5 years. Data available for credit cycles are within the range from 5 up to 20 years, it indicates that credit cycles are generally three or four times longer than the business cycles, which are in medium 5 years long. Another approach suggests the setting of business cycles. by comparing the length of credit and Let s assume four options: The lenght of credit and business cycles is the same The lenght of the credit cycle is two times longer comparing to business cycle The lenght of the credit cycle is three times longer comparing to business cycle The lenght of the credit cycle is four times longer comparing to business cycle (Borio et al., 2010: 28) The empirical analysis done by Borio, Drehmann, Gambacorta, Jimenez and Trucharte (2010: 28 30) assumes that works well in case of excessive credit period determination in a private sector. 21

32 current US$ Hodrick-Prescott filter and CEE countries First of all we will briefly describe Central and Eastern European countries (shortly CEE countries). In this analysis we will discuss in detail following group of nations: Czech Republic, Estonia, Hungary, Latvia, Lithuania, Poland and Slovak Republic. The main reason for picking the CEE countries is that this region recorded a significant credit boom in private sector, before the global financial crisis, especially Baltic States reported significantly excessive credit growth (Backé et al., 2006: 5). Financial position of CEE countries CEE countries are considered as transition economies. All CEE countries are members of European Union, but only two of them are members of Euro zone and so: Estonia and Slovakia. Graph Comparison of GDP per capita European Union OECD members World Czech Republic Estonia Hungary Latvia Lithuania Poland Slovak Republic (Source: World Bank data database) As we can see in the graph above economic output of CEE countries is similar, however comparing to an average between OECD members or European Union as a whole they are still significantly behind. 22

33 Using just credit-to-gdp ratio in the case of CEE countries could be misleading, because in this type of countries can rapid excessive credit growth be only a reason of convergence to levels of advanced economies (Hilbers et al., 2005: 10). Here (Graph 5.) we can clearly see the difference in private-to-gdp ratio between CEE countries, European Union, World and OECD members. Although CEE countries are approaching, still the difference is significant. Graph 5 % of GDP Comparison of the Domestic credit to private sector OECD members World European Union Estonia Latvia Hungary Lithuania Czech Republic Slovak Republic Poland (Source: World Bank data database) Setting the parameter Although the technical literature usually suggests to set for computations in private sector, we can face few issues in case of CEE countries. According to the proposal of Borio et al. (2010: 28 30), works very well for countries with long data history, as this method was estimated on long time series of advanced nations. Good example is United Kingdom where we can use quarterly data even from 1960, what in case of CEE countries is not possible. As we described before, determining of lambda is significantly dependent on length of time series data. 2 Problem lies then in shortage of time series data availability. In the early 90s started process of transition of economic systems from the socialistic planned economy 2 see part Determination of λ parameter 23

34 in order to set up free market in the CEE countries and thus converge to the advanced economies from Western Europe. CEE countries experienced a difficult economic period after era of socialism. Publication of statistical data was in many cases limited or even prohibited. Problem of data limitation was partially solved by estimation what on the other hand lead to biased results. Also influence of bad loan inherited from old regime and privatization in 90. years resulted in inaccurate statistical indicators in this period (Takata, 2005: 1 15). Therefore quality data for CEE countries are generally available just from one decade ago. In our analysis we have chosen Czech Republic as the representing country of the CEE countries. The data used in calculations were obtained mainly from the International Financial Statistics (IMF) database of the International Monetary Fund (IMF). The main reason of choosing was that it provides sufficiently long quarterly data in long time series. However it applies only for the advanced countries (United Kingdom, Germany or France), but in case of CEE countries the compact data are available just from 90. years. As another source of data we used World Bank s database. Calculation of countercyclical capital buffers As it is stated in BCBS (2010a: 21 23), the countercyclical capital buffer proposal, there are three steps in determining the buffer, based on credit-to-gdp ratio. First step is to calculate the aggregate private credit-to-gdp ratio. Second part is to determine the credit-to-gdp gap and last part is transformation of these data into a countercyclical capital buffer. I. Calculation of the aggregate private credit-to-gdp ratio We consider time series data with period t for each country. The credit-to-gdp ratio is calculated as follows: [2] 24

35 Where is a nominal domestic GDP in period t. is an indicator that measures nominal credit of the private, non-financial sector in a given period t. The empirical analysis states that a broad definition of credit is a better predictor of stressful period than a narrow one. This means that there should be included not only domestic credit, but also credit provided by international banks and non-bank institutions (BCBS, 2010b: 10). II. Calculation of the credit-to-gdp gap Calculation of the credit-to-gdp gap is based on comparing credit-to-gdp ratio with its long term trend. The gap can be easily calculated as: [3] Where is above mentioned credit-to-gdp ratio in period t. is long term trend in given period achieved by the HP filter. The smoothing parameter lambda is generally set to (Borio et al., 2010: 28 30) However setting lambda is up to each jurisdiction, we have already discussed. III. Transformation of the credit-to-gdp ratio into a countercyclical buffer The real size of the buffer is expressed as percentage of risk-weighted assets. The buffer is equal zero when the credit-to-gdp gap is below a certain lower threshold (L) level. The size of the buffer is linearly increasing with rising values of credit-to-gdp gap until certain level called upper threshold (H) where it approaches its maximum. BCBS analysis has discovered that thresholds at levels L=2 and H=10 provide sensible and robust specification based on historical experience with financial crises. Indeed it depends not only on adjusting the threshold levels, but also on selecting of the smoothing parameter lambda and the length of the available time series data. We have three situations for certain credit-to-gdp gap values: 25

36 ; where is linearly distributed in interval This example demonstrates how is the countercyclical capital buffer calculated in case of United Kingdom from the period right before the financial crisis in Table 2: Calculation of the countercyclical capital buffer for United Kingdom Year GDP Adjusted, Nominal, Billions Claims on Private Sector, Billions Credit-to- GDP in % gap in % λ=20000 trend in % λ=20000 gap in % λ= trend in % λ= Buffer as % of RWA (λ=400000) 2003 Q1 1089, , ,538-3, ,412-7, , Q2 1106, , ,065-4, ,846-7, , Q3 1121, , ,414-3, ,310-6, , Q4 1139, , ,582-4, ,804-6, , Q1 1155, , ,122-2, ,326-4, , Q2 1171, , ,813-3, ,874-4, , Q3 1186, , ,509-1, ,448-3, , Q4 1202, , ,336-2, ,044-3, , Q1 1215, , ,420-2, ,662-2, , Q2 1228, , ,201-2, ,299-2, , Q3 1240, , ,922-2, ,952-1, , Q4 1254, , ,043-0, ,621 0, , Q1 1272, , ,442 4, ,303 5, ,769 1, Q2 1290, , ,899 3, ,995 5, ,960 1, Q3 1309, , ,966 3, ,696 5, ,153 1, Q4 1328, , ,819 3, ,402 6, ,345 1, Q1 1345, , ,146 8, ,114 11, ,539 2, Q2 1365, , ,143 8, ,828 12, ,732 2, Q3 1385, , ,583 12, ,544 16, ,926 2, Q4 1405, , ,790 13, ,261 18, ,119 2,5 (Source: IMF IFS, author s calculations) The data used in this calculation were acquired in International Monetary Fund s statistical database. We are working with data between periods in order to make our estimations as precise as possible. As variables for credit-to-gdp indicator was used Claims on private sector as a credit indicator and nominal value of GDP in given period. This variable slightly underestimates true value of the credit, because nonbank providers are not included. However for our computations it is not crucial, whereas the financial system in Czech Republic is mostly bank-based, therefore the 26

37 1970 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q3 % most of loans is provided by the bank institutions. According to Czech National Bank s Financial Stability Report (2007: 46) bank institutions represented 74, 2% of the financial sector assets. Long-term trend was obtained by HP filter in statistical program STATA with different levels of lambda (25000, ). We selected especially these two levels of lambda to clearly demonstrate their differences in computation. Individual thresholds of build-up phase are distinguished by different colors. As 3 percentage of Risk Weighted Assets (RWA) we considered 2, 5% as was recommended by BCBS (2010a: 2). National authorities can adjust this upper maximum as it is best for them. It will cover only their national banks, but international institutions will not be included in this regime. The estimation by HP filter based on data available before the crisis using suggested value λ= showed us that build-up phase should started at maximum level in the first quarter of This result is considered as the optimum. Here is a comparison of different levels of λ: Graph 6 30 United Kingdom, HP filter prediction gap 1600 gap gap gap (Source: IMF IFS, author s calculations) Variations between different lambdas are significant. Assuming that λ= is an appropriate parameter for determining build-up phase in case of United Kingdom, 3 See also Transformation of the credit-to-gdp ratio into a countercyclical buffer at previous page. 27

38 1970 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q4 % the build-up phase at maximum (means 10% of Risk Weighted Assets), should be released at the first quarter of Graph 7&8 Credit-to-GDP in % Comparison Credit-to-GDP and its HP trend Comparison of the Gap and the Buffer Credit-to-GDP in % trend gap Buffer as % of RWA (λ=400000) (Source: IMF IFS, author s calculations) In the graph 7 we can see the comparison of the credit-to-gdp ratio and its long-term trend calculated by HP filter. Periods of financial strains are characterized by major deviations from its long-term trend. The graph 8 above shows us demonstration how countercyclical capital buffers works. When the credit-to-gdp gap exceeds 2%, buffer is automatically created. In period when values of the gap are negative, we can see that buffer is equal zero. Disadvantages of Credit-to-GDP gap indicator As we mentioned before, using Credit-to-GDP indicator with HP filter can cause several problems when we apply it for CEE countries. First thing worth of consideration is value of smoothing parameter λ, which is generally designed to be , what is probably in case of CEE countries considerably overestimated. 28

39 As we have already mentioned CEE countries tend to converge in values of credit-to-gdp ratio to developed nations as for example United Kingdom. 4 It may happen that our results of estimation can be biased, because of upper mentioned convergence. Here is comparison between the CEE countries and United Kingdom as a representative of developed countries in the first quarters of 2000 and 2007: Graph 9 2 1,5 1 0,5 0 Comparison of the credit-to-gdp growth 2007 Q Q1 (Source: IMF IFS, author s calculation) In the graph above the differences between advanced countries represented e.g. by United Kingdom or Denmark and CEE countries are still relatively apparent. It is worth to note also the significant growth between these two periods, especially in case of the CEE countries Latvia, Lithuania and Estonia, meanwhile the rest of them appears to be relatively stable. Another disadvantage of the estimator using credit-to-gdp gap is its ambiguous representation of the results in previous periods. The HP filter significantly distorts results from periods before the current one. It means that looking back to history we are sometimes unable to locate previous periods of financial strains, because numbers depending on whole data period are biased. Here is an comparison of credit-to-gdp gap after whole period until 2012 and only with data until 2000: 4 Example of United Kingdom is suitable because of two reasons. One is historicaly sufficient long time series data and second is a possibility to compare results with other literature (see [Borio et al., 2010]) 29

40 1993 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q3 % % Graphs 10& Credit-to-GDP gap, Czech Republic until 2000 gap gap gap Credit-to-GDP gap, Czech Republic until 2012 gap gap gap (Source: IMF IFS, author s calculation) First considerable difference is that credit-to-gdp gaps are in first picture almost the same, copying each other, though in second one (shadow zone denotes the same period), we can see different fluctuations. This might be caused by not sufficient observations in the first case. Other difference is when we compare values in one period. For example results for λ= in 1993 Q4 in the first picture is 1,33 on the other hand second graph shows us for the same period 7,45, what is considerably different. Determination of the excessive level of credit The main problem in creating an excessive credit level indicator is that it is not straightforward to estimate which level of credit can pose a given country into the risk. There are more approaches available. Traditional approach is to use the HP filter. The HP filter obtains the trend from a time series of credit-to-gdp ratio. Comparing the actual ratio with its long term trend obtained by HP filter we can determine whether is the level of credit in economy excessive or not. The advantage of using the HP filter is that it tends to give higher weights to the more late observations what acts more effectively in case of structural breaks (BCBS, 2010b: 13). 30

41 In order to compare the results of the HP filter with other indicators and ensure us about the correctness of the predictions of excessiveness we proposed another set of indicators for excessiveness. Another reason is that credit-to-gdp ratio can be in case of CEE countries misleading as they can only converge to levels of advanced economies (Hilbers et al., 2005: 10). Although there are many different methods we, as many others before us: Backé et al. (2006) or Kiss et al. (2006), consider credit growth as a one of key variables in estimation of excessiveness, thus it can be also good estimator of coming recessions and financial strains. In our indicators we used as a variable nominal credit in given countries, specifically Claims on private sector. Data were obtained from IMF IFS database in quarterly period. This variable omits loans which are provided by other non-bank institutions or from abroad, therefore it underestimates real value of private credit, but this is in our case not crucial. We have already mentioned before that financial system in Czech Republic is mostly bank-based (CNB, 2007: 46). Core of these indicators lies in a calculation of the nominal quarterly credit growth between two adjacent periods. For estimation of excessive credit growth we applied two different methods to confirm their correctness. Average credit growth indicator The first method proposed is based on determination of quarterly credit growth for all periods and calculation of the average of all obtained non-negative values. 5 Threshold (denoted as ω) for determination of state of excessiveness is different for every country as it is equal to the average of obtained values multiplied by given parameter (denoted as ρ). The parameter ρ is fixed for all countries and depends on the length of the time series used. The longer sample of observations we have, the higher parameter we need to set. 6 The period of excessiveness is defined as the period when quarterly growth between two periods is higher than given threshold value. The threshold ω was set as an average of all growth values multiplied by parameter ρ: 5 As we will mention later, this condition is hepful in cases where given countries have experienced period of consolidation, it does not influence other countries. 6 Comparison of different levels of the parameter ρ is demonstrated in following the Graph 12. Additional comparison based on times series length is incuded in Appendix C. 31

42 2000 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q3 [4] Function used to determine excessiveness is: [5] Main advantage of this method is that, it is not so biased by period of consolidation, which was typical for some of the CEE countries. Decrease in credit provided during the end of 90. years and also beginning of 2000 was due to the fact that banks were cleaning their portfolios and selling bad loans to the consolidating agencies in order to solve the poor management of credits during the socialistic era (Vencovský, 1999: 509). Among the countries having problems with consolidation are mainly Czech Republic, Slovakia and Baltic countries, but not in such an extent. We found it out on the basis of credit statistics. 7 However negative values in credit especially in case of Czech Republic and Slovakia make averages distorted. Therefore in order to obtain results with the highest accuracy, these negative values were not included. We will discuss this problem also later, especially in the case of the Average trend credit indicator. However this method has the same problem with data interpretation as the Credit-to-GDP indicator and so: As the threshold values is depended on all observations, from a long-time horizon is hard to determine past periods of excessiveness. Therefore we have to determine it recursively as in the case of HP filter method. Graph 12 The average credit growth indicator in United Kingdom growth ω with ρ=1 ω with ρ=1,5 ω with ρ=2,38 7 For further information see Appendix B. 32

43 (Source: IMF IFS, author s calculation) Parameter was determined after empirical analysis, supposing to highlight just significant differences. As we can see, lower parameter has caused frequently occurring period of excessiveness, on the other hand too high parameter will not give us any results neither. By gradually setting of parameter we reached the result that for this length of time series is appropriate to set it from the interval. In order to observe coming periods of excessiveness in advance, we recommend to run this indicator on several different levels of the parameter from the given interval. Table 3: The Average credit growth indicator s results comparison for other countries period Estonia Latvia Lithuania Czech Republic Hungary Poland Slovakia 2005 Q1 ok ok ok ok ok ok ok ok 2005 Q2 ok ok ok ok ok ok ok ok 2005 Q3 ok ok ok ok ok ok ok ok 2005 Q4 ok ok excessive ok ok ok ok ok 2006 Q1 ok ok ok ok ok ok ok ok 2006 Q2 ok ok ok ok ok ok ok ok 2006 Q3 ok ok ok ok ok ok ok ok 2006 Q4 excessive excessive ok ok ok ok ok ok 2007 Q1 ok excessive ok ok ok ok ok ok 2007 Q2 ok ok ok excessive ok excessive ok ok 2007 Q3 ok ok excessive ok ok excessive ok ok United Kingdom 2007 Q4 ok ok excessive ok ok ok ok ok (Source: IMF IFS, author s calculation) In the previous table is illustration of how the Average credit growth indicator works for a selected group of countries. This table shows only important periods right before crisis, although computation is based on data from period 2000 until the end of We used parameter. As we can see Baltic countries according to our indicator have experienced periods of credit booms, also there is sign of excessiveness in Czech Republic and Poland. For example United Kingdom does not show any signs of excess, however calculation recursively until the period 2007 Q1 will show us excessiveness in this point as well. 33

44 2000 Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q Q3 Average trend credit indicator In the second method we applied credit trend as a comparative variable to the credit growth. Trend was calculated as a function of previous observations including the current value. Periods before credit booms has typically recorded unusual increase in credit growth and therefore are significantly different from their long-term trend. Again we work with the hypothesis that only the most significant differences between credit growth and its trend can lead to financial stressful period in the economy. In order to properly distinguish the periods of credit booms and period of common fluctuations we determined threshold limit as in the first indicator. Let denote threshold for the Average trend credit indicator, then threshold limit was determined as follows: [6] Where is a parameter and as in the first case depends on the number of observations. As a period of the excessive credit growth is indicated the point where credit growth between two adjacent years prevails the threshold value in given period. [7] Graph 13 The average trend credit indicator in United Kingdom growth θ with σ=1,3 θ with σ=1,65 (Source: IMF IFS, author s calculation) 34

45 The same procedure as before was used to determine the parameter and in this case we recommend to use it from the interval. We draw attention once again that this is working in both cases only for this specific length of time series. Table 4: The Average trend credit indicator s results comparison for other countries period Estonia Latvia Lithuania Czech Republic Hungary Poland Slovakia 2003 Q2 ok ok ok ok ok ok ok ok 2003 Q3 ok ok ok ok ok ok ok ok 2003 Q4 ok ok ok ok ok ok ok ok 2004 Q1 ok ok ok ok ok excessive ok ok 2004 Q2 ok ok ok ok excessive ok excessive ok 2004 Q3 ok ok ok ok ok ok ok ok 2004 Q4 ok ok ok ok excessive ok ok ok 2005 Q1 ok ok ok ok ok excessive excessive ok 2005 Q2 excessive ok excessive excessive ok ok excessive ok 2005 Q3 excessive ok excessive ok excessive excessive excessive ok United Kingdom 2005 Q4 ok ok excessive ok excessive excessive ok excessive 2006 Q1 excessive ok ok ok ok excessive ok excessive 2006 Q2 excessive ok ok ok excessive excessive ok ok 2006 Q3 ok ok ok ok ok excessive ok ok 2006 Q4 excessive excessive ok ok ok excessive ok ok 2007 Q1 ok ok ok ok ok excessive ok excessive 2007 Q2 ok ok ok excessive excessive excessive ok ok 2007 Q3 ok ok excessive ok excessive excessive ok excessive 2007 Q4 ok ok ok ok excessive ok excessive ok (Source: IMF IFS, author s calculation) The previous table shows results of the Average trend credit indicator for other countries. We used data only from 2003 until 2007, because we want to avoid period of consolidation in some of the countries. For computation we used parameter. As we can see this indicator comparing to previous one seems to be more restrictive. For Baltic countries and Czech Republic results seem reasonable. Slovakia has reported period of excessiveness during the year 2005, what is probably caused by huge decline in credit in the first quarter of 2004 when it recorded a great fall of credit from 18391,4 to -428,9 millions of Slovak crowns. However problem arises with Hungary and Poland. In this case we have to define special conditions for the proper functioning of given indicator. As it is based on 35

46 trend estimation, period with significant credit loses underestimates trend variable for several following periods what makes this indicator biased. To ensure its functionality we can use it only on time series with steadily growing credit with only small decreases, what is in the case of Hungary and Poland not fulfilled. On the other hand, United Kingdom is a good example of an appropriate adept for this method, looking on the given conditions. Nevertheless the main advantage of this method is that it does not depend on the length of time series we use, because threshold is computed from data until the given period only. When using the Average credit growth indicator we have problem that longer time series distorts results from the past. The explanation is following: Amount of credit is generally an increasing function and above-mentioned indicator is using average as a threshold measure. Therefore higher values in the last years can overestimate the threshold and thus it will conceal past periods of potential excessiveness. Following table demonstrates comparison of the results for the Average credit growth indicator, Average trend credit indicator and Credit-to-GDP indicator in the case of United Kingdom. Data used are from 2000 until 2008 in quarterly period. Table also shows different levels of the parameter ρ and σ and their results. 36

47 Table 5: Calculation of excessiveness by the Average credit growth indicator and The average trend credit indicator for United Kingdom period Claims on Private Sector 2000 Q1 1155,655 growth The Average credit growth indicator excess ρ=1 excess ρ=2,35 The Average trend credit indicator excess σ=1,5 excess σ=1, Q2 1196,596 40,941 ok ok ok ok ok 2000 Q3 1223,012 26,416 ok ok ok ok ok 2000 Q4 1254,261 31,249 ok ok ok ok ok 2001 Q1 1301,66 47,399 ok ok ok ok ok 2001 Q2 1312,762 11,102 ok ok ok ok ok 2001 Q3 1341,603 28,841 ok ok ok ok ok 2001 Q4 1367,753 26,15 ok ok ok ok ok 2002 Q1 1388,527 20,774 ok ok ok ok ok 2002 Q2 1427,402 38,875 ok ok ok ok ok 2002 Q3 1445,694 18,292 ok ok ok ok ok 2002 Q4 1479,428 33,734 ok ok ok ok ok 2003 Q1 1509,897 30,469 ok ok ok ok ok 2003 Q2 1538,277 28,38 ok ok ok ok ok 2003 Q3 1586,289 48,012 excessive ok excessive ok ok 2003 Q4 1624,635 38,346 ok ok ok ok ok 2004 Q1 1688,749 64,114 excessive ok excessive ok ok 2004 Q2 1719,809 31,06 ok ok ok ok ok 2004 Q3 1774,429 54,62 excessive ok ok ok ok 2004 Q4 1807,598 33,169 ok ok ok ok ok 2005 Q1 1852,256 44,658 ok ok ok ok ok 2005 Q2 1893,616 41,36 ok ok ok ok ok 2005 Q3 1934,505 40,889 ok ok ok ok ok 2005 Q4 1994,865 60,36 excessive ok ok ok ok 2006 Q1 2104, ,661 excessive ok excessive excessive ok 2006 Q2 2153,974 49,448 excessive ok ok ok ok 2006 Q3 2200,149 46,175 ok ok ok ok ok 2006 Q4 2256,205 56,056 excessive ok ok ok ok Credit-to- GDP gap excess λ= Q1 2369, ,747 excessive excessive excessive excessive excessive 2007 Q2 2431,827 61,875 excessive ok ok ok excessive 2007 Q3 2544, ,475 excessive excessive ok ok excessive 2007 Q4 2625,88 81,578 excessive ok ok ok excessive (Source: IMF IFS, author s calculation) Comparing our results with Credit-to-GDP method using HP filter we can conclude that our estimates bring similar results. Credit-to-GDP indicator denotes as 37

48 the period of excessiveness in United Kingdom the first quarter of 2007, what we find out also in both of our cases. On higher levels of parameters we find out also the first quarter of 2006 behave like period of excessiveness and also some periods around 2004, but this can be reduced by lower setting of the parameters and it serves like a good indicator of coming credit booms. As we have already mentioned before, some of the CEE countries experienced period of consolidation. Therefore in the case of Czech Republic, Slovakia and all of the countries where we recorded significant, but temporary decline in credit in some years, we can still use this method, but only with time series adjusted for credit drops. 8 This means that we apply this indicator on the data after the period of consolidation. In our previous example, in case of United Kingdom, there was no problem, however in case of Czech Republic or Slovakia we had to start with period 2003 Q1 or we have to ignore periods of excessiveness in years where these conditions are not met. Comparison of the results In this section we compare and discuss our previous results with financial stability reports from given countries. We focus mainly on the period just before The Great financial crisis in We have already mentioned that United Kingdom is one of the representative countries on which was built Credit-to-GDP indicator. The results of our indicators and also Credit-to-GDP gap method assumed that the first quarter of 2007 is an excessive credit period. The financial stability report (2007: 16, ) from Bank of England supports our claim by: Growth in loans to finance leveraged buyouts (LBOs) has been particularly strong and the proportion of sub-investment grade debt in global syndicated loan issuance exceeded 50% in 2007 Q1. According to the Average credit growth indicator in Slovakia, it has not experienced excessive period, on the other hand the Average trend credit indicator 8 Difference between period of consolidation and Hungary and Poland is that in the case of consolidation there is only temporal period of credit drops which can be excluded, while in Hungary and Poland we deal with this problem along the whole time series. 38

49 proclaimed the fourth quarter of 2007 as period of excessiveness. National Bank of Slovakia stated in their financial report (2007: 5 6, ):...the trend of increasing credit risks in the banking sector will be maintained, however, the rate of increase will be slower than in the previous years. And also: In 2007, several banks recorded a significant increase in loans..., although most banks have not yet been found to be exposed to a significant impairment of retail loans... Assuming no excessiveness, the excessive growth signalized by our indicator at the end of 2007 could be caused by inadequately setting of parameter. In case of Poland we reported higher increase in credit in the second and third quarter of However Average trend credit indicator is not working here due to unfulfilled preconditions. Discussing our results with The Financial stability report (2007: 28, ) from National Bank of Poland we find out: In June 2007 a relatively high growth in the value of irregular loans was recorded at some small banks that focus on the retail market., However according to Credit-to-GDP indicator Poland does not indicate any signs of excess. Questionable in this case then is whether our indicators show reasonable results for Poland or whether the Credit-to-GDP method is in this case appropriate. For Hungary, similarly as for Slovakia, Average credit growth indicator has not observed any period of excessiveness; due to same reason as in the Poland Average trend credit indicator does not function in this case. Consulting Financial stability report (2007: 29, 33; ): in 2006, the growth rate of long-term loans also tapered off...by the end of investment financed from loans is gradually gaining ground. In case of Baltic countries we observed excessive period by both of indicators in Estonia and Latvia mainly at the end of Lithuania indicated excessiveness in the second half of We discussed the study about the credit growth in Eastern Europe 39

50 from Backé, Égert and Zumer (2007: 73) which states that in private sectors in 2006 were Credit-to-GDP levels in case of Lithuania close to the lower bound of the estimated equilibrium ranges as in the case of Slovakia. Although Lithuania did not signal any period of excessiveness in 2006, at the end of 2007 indicated growth in credit as our indicators had predicted. Comparing to Lithuania, level in Estonia was more elevated and the increase was the most notable in Latvia. Speaking about the Czech Republic both of our methods confirmed that in the second quarter of 2007 was potential period of excessive credit. Consulting with the Financial stability report (2007: 28, 29; ): The existing studies analyzing the equilibrium level of debt of the private sector (as measured, for example, by the ratio of loans to the private sector to GDP) confirm that the rise in debt in the Czech Republic is in line with the country's overall economic growth. This means that there is no "excessive" growth in lending that might mask certain risks to the future stability of the financial system. Questionable now is whether the Czech Republic has experienced period of excessiveness in the second quarter of 2007 or not. According to Backé et al. (2007: 73) this period of the second quarter of 2007 was also characterized as marginally lying on the boundary signalizing excessiveness. The reason why we consider in our observations 2007 Q2 as period of excessive credit growth can be explained by the fact that we use more conservative approach and therefore we set our parameter to be more sensitive. Bank capital to assets ratio Another measure which is worth to mention while speaking about credit booms is Bank capital to assets ratio. Its definition is following: Bank capital to assets is the ratio of bank capital and reserves to total assets. Capital and reserves include funds contributed by owners, retained earnings, general and special reserves, provisions, and valuation adjustments. Total assets include all nonfinancial and financial assets. (World Bank, ) In other words this ratio is a criterion which determines the coverage consisting of capital against a credit risk. The following table shows the comparison of the bank capital to assets ratio between 2000 and We used the same periods as we are testing our indicators on. 40

51 Graph 14 Comparison of Bank capital to assets Czech Republic Estonia Hungary Latvia Lithuania Poland Slovak Republic United Kingdom (Source: World Bank, author s calculation) As we can see Baltic countries and United Kingdom have recorded a decrease in the bank capital to asset ratio (values on y-axis) between these two periods. Decline in this ratio can be caused by two ways. Either they recorded significant decrease in bank capital and reserves or increase in assets, respectively partially both of them. However based on the results of previous indicators we know that these countries have experienced periods of excessiveness. Although countries like Czech Republic, Slovakia, Hungary and Poland based on our indicators have also recorded some periods of excessiveness, but probably not in such an extent. Therefore we are proposing hypothesis that these significant decreases in the bank capital to assets ratio can be additional measures which can be used to verify previous results. Appropriate lambda for CEE countries This section is focused on the discussion of properties of the suggested parameter λ= in the case especially of Czech Republic, comparison of the results with our indicators and proper setting of the lambda parameter. Results from Czech Republic using λ= are as follows: 41